Sphygmomanometer, blood pressure measurement method, blood pressure measurement program, learning model construction method, and learning model construction program

Abstract

The present invention enhances the accuracy of estimation of a blood pressure value using a mathematical model. This sphygmomanometer comprises: a control unit that stops pressurization when a first pulse wave acquired while a measurement site is gradually pressurized using a cuff satisfies a prescribed pressurization stop condition; a generation unit that generates, on the basis of the first pulse wave acquired before the prescribed pressurization stop condition is satisfied, a first mathematical model of a predicted envelope of a second pulse wave expected to be obtained when pressurization is continued; a correction model in which, for a subject, a feature amount relating to the first pulse wave acquired before the prescribed pressurization stop condition is satisfied is set as an explanatory variable, and information relating to a parameter of a second mathematical model of a measured envelope of a third pulse wave continuously acquired even after the prescribed pressurization stop condition is satisfied is set as an objective variable; and a correction unit that corrects the first mathematical model using information output by the correction model in response to input of the feature amount to the correction model. The sphygmomanometer estimates a blood pressure value of a measurement subject using the first mathematical model corrected by the correction unit.

Description

Sphygmomanometer, blood pressure measurement method, blood pressure measurement program, learning model construction method and learning model construction program

[0001] The present invention relates to a sphygmomanometer, a blood pressure measurement method, a blood pressure measurement program, a learning model construction method, and a learning model construction program.

[0002] Health status is grasped through blood pressure measurement. Patent Document 1 describes estimating a mathematical model of an envelope of a pressure pulse wave and calculating a blood pressure value using the estimated mathematical model.

[0003] U.S. Patent No. 11179050 Specification

[0004] When estimating a mathematical model of the envelope of a pressure pulse wave, an error may occur between the envelope represented by the estimated mathematical model and the actual envelope. Further, when the acquired pressure pulse wave is short, the envelope of the pressure pulse wave also becomes short. When the envelope is short, the error caused by the estimation of the mathematical model also becomes larger. Such an error between the mathematical model and the actual envelope reduces the estimation accuracy of the blood pressure value using the mathematical model.

[0005] One aspect of the disclosed technology aims to provide a sphygmomanometer, a blood pressure measurement method, a blood pressure measurement program, a learning model construction method, and a learning model construction program that can improve the estimation accuracy of blood pressure values using a mathematical model.

[0006] One aspect of the disclosed technology is exemplified by the following blood pressure monitor: This blood pressure monitor includes a sensor that detects the pressure of a cuff for compressing the part of the person to be measured; a pulse wave acquisition unit that acquires a pulse wave from the output signal of the sensor; a control unit that stops pressurizing when a first pulse wave acquired while gradually pressurizing the part of the person to be measured using the cuff satisfies a predetermined pressurization stop condition; a generation unit that generates a first mathematical model of the predicted envelope of a second pulse wave that is expected to be obtained if pressurization is continued, based on the first pulse wave acquired up to the time the predetermined pressurization stop condition is met; and for the subject, the first pulse wave acquired up to the time the predetermined pressurization stop condition is met The system comprises: a correction model that uses features related to pulse waves as explanatory variables and information relating to the parameters of a second mathematical model of the measured envelope of a third pulse wave, which has been continuously acquired even after the predetermined pressurization stop condition has been met, as the objective variable; a correction unit that inputs the features into the correction model to acquire the information output by the correction model and corrects the first mathematical model using the acquired information; an estimation unit that estimates the blood pressure value of the subject using the first mathematical model corrected by the correction unit; and an output unit that outputs the blood pressure value estimated by the estimation unit.

[0007] With this blood pressure monitor, the generated first mathematical model is corrected by the correction unit, thereby increasing the accuracy of blood pressure estimation as much as possible. In other words, this blood pressure monitor can correct errors that may be included in the first mathematical model of the predicted envelope, thereby increasing the accuracy of blood pressure estimation using the mathematical model. Here, the above information may be the error between the parameters of the first mathematical model and the parameters of the second mathematical model, or it may be the parameters of the second mathematical model.

[0008] Here, the information relating to the first pulse wave may be the amplitude of the first pulse wave, or it may be the ratio of the area of ​​the positive region to the area of ​​the negative region of the first pulse wave. In other words, this blood pressure monitor can appropriately select the above information based on the accuracy of blood pressure estimation, the computational load on the blood pressure monitor, etc.

[0009] The disclosed technology can also be understood from the perspectives of blood pressure measurement methods, blood pressure measurement programs, learning model construction methods, and learning model construction programs.

[0010] According to the disclosed technology, it is possible to improve the accuracy of blood pressure calculations using mathematical models.

[0011] Figure 1 is a diagram showing an example of a blood pressure monitor according to the embodiment. Figure 2 is a diagram showing an example of the processing block of the blood pressure monitor body according to the embodiment. Figure 3 is a diagram schematically showing the generation of a mathematical model of the predicted envelope by the generation unit in the embodiment. Figure 4 is a diagram showing the relationship between the value of index α by (Equation 2) and the amplitude of the envelope. Figure 5 is a diagram explaining the feature quantity of (11) in the embodiment. Figure 6 is a diagram explaining the feature quantity of (12) in the embodiment. Figure 7 is a diagram explaining the feature quantity of (19) in the embodiment. Figure 8 is a diagram schematically showing the correction of the mathematical model showing the predicted envelope E2 by the correction unit according to the embodiment. Figure 9 is a diagram schematically showing the construction of the correction model according to the embodiment. Figure 10 is a diagram showing an example of the processing flow of the blood pressure value measurement process by the blood pressure monitor according to the embodiment.

[0012] <Application Examples> Examples of applications of the present invention will now be described. An example of an application of the present invention is the blood pressure monitor 1 illustrated in Figure 1. The blood pressure monitor 1 acquires a pressure pulse wave while gradually applying pressure to the measurement site of the person being measured with the cuff 16. Then, when the envelope of the acquired pressure pulse wave satisfies a predetermined pressure stop condition, the blood pressure monitor 1 stops applying pressure with the cuff 16.

[0013] The blood pressure monitor 1 generates a mathematical model of the predicted envelope E2 (see Figure 3) that is expected to be obtained if pressurization is continued, based on the pressure pulse wave acquired until the predetermined pressurization stop condition is met. However, the generated mathematical model may contain errors during the modeling process. Furthermore, if the blood pressure value is estimated based on a mathematical model containing errors, the estimated blood pressure value will also contain errors. In this application example, the following configuration is adopted to suppress the decrease in the accuracy of blood pressure value estimation due to such errors.

[0014] The blood pressure monitor 1 includes a correction model 105 (see Figure 2) in which the characteristic quantities related to the pressure pulse wave P1 (see Figure 3) acquired before the predetermined pressurization stop condition is met are used as explanatory variables, and information related to the parameters of the mathematical model of the measured envelope E11 (see Figure 9) of the pressure pulse wave P2 (see Figure 9) that was continuously acquired even after the predetermined pressurization stop condition was met is used as the objective variable. The blood pressure monitor 1 then uses the information acquired by inputting the characteristic quantities of the pressure pulse wave P1 into the correction model 105 to correct the mathematical model of the envelope E1, and estimates the blood pressure value using the corrected mathematical model. Therefore, the blood pressure monitor 1 in this application example can improve the accuracy of blood pressure value estimation using a mathematical model.

[0015] <Embodiments> Embodiments will now be described. Figure 1 is a diagram showing an example of a blood pressure monitor 1 according to an embodiment. The blood pressure monitor 1 is a device that measures the blood pressure of a person by wrapping a cuff 16 around the area to be measured and compressing the area with the cuff 16. The area to be measured can be, for example, the upper arm. The blood pressure values ​​to be measured include, for example, SBP and DBP.

[0016] Figure 1 shows an example of the hardware configuration of the blood pressure monitor 1. The blood pressure monitor 1 comprises a blood pressure monitor body 10 and a cuff 16. The cuff 16 is formed in a strip shape and has a bag inside into which air is supplied. When the cuff 16 is wrapped around the body part to be measured and air is supplied to the bag inside the cuff 16, the pressure exerted by the cuff on the body part to be measured increases. Conversely, when the cuff 16 is wrapped around the body part to be measured and air is drawn in from the bag inside the cuff 16, the pressure exerted by the cuff on the body part to be measured decreases.

[0017] The blood pressure monitor unit 10 is an information processing device comprising a Central Processing Unit (CPU) 11, a main memory unit 12, an auxiliary memory unit 13, a display 14, a connection unit 15, a cuff 16, and a connection bus B1. The CPU 11, main memory unit 12, auxiliary memory unit 13, display 14, and connection unit 15 are interconnected by the connection bus B1.

[0018] The CPU 11 is also called a microprocessor unit (MPU) or processor. The CPU 11 is not limited to a single processor and may be in a multiprocessor configuration. A single CPU 11 connected via a single socket may also have a multicore configuration. At least a portion of the processing performed by the CPU 11 may be performed by a processor other than the CPU 11, such as a dedicated processor like a Digital Signal Processor (DSP), Graphics Processing Unit (GPU), numerical processor, vector processor, or image processing processor. At least a portion of the processing performed by the CPU 11 may also be performed by an integrated circuit (IC) or other digital circuit. Furthermore, at least a portion of the CPU 11 may include analog circuits. Integrated circuits include Large Scale Integrated Circuits (LSIs), Application Specific Integrated Circuits (ASICs), and Programmable Logic Devices (PLDs). PLDs include, for example, Field-Programmable Gate Arrays (FPGAs). The CPU 11 may be a combination of a processor and an integrated circuit. Such combinations are called, for example, a microcontroller unit (MCU), a System-on-a-chip (SoC), a system LSI, or a chipset. In the blood pressure monitor 1, the CPU 11 deploys a program stored in the auxiliary storage unit 13 to the work area of ​​the main storage unit 12 and controls peripheral devices through program execution. This allows the blood pressure monitor 1 to perform processing that matches a predetermined purpose. The main memory unit 12 and the auxiliary memory unit 13 are recording media that can be read by the CPU 11.

[0019] The main memory unit 12 is exemplified as a memory unit that is directly accessed by the CPU 11. The main memory unit 12 includes Random Access Memory (RAM) and Read Only Memory (ROM).

[0020] The auxiliary storage unit 13 stores various programs and data on a recording medium in a read-write manner. The auxiliary storage unit 13 is also called an external storage device. The auxiliary storage unit 13 stores the Operating System (OS), various programs, various tables, etc. The auxiliary storage unit 13 may, for example, be part of a cloud system, which is a group of computers on a network.

[0021] The auxiliary storage unit 13 is, for example, an Erasable Programmable ROM (EPROM), a Solid State Drive (SSD), a Hard Disk Drive (HDD), etc.

[0022] The display 14 displays data processed by the CPU 11 and data stored in the main memory 12. The display 14 is, for example, a Liquid Crystal Display (LCD), a Plasma Display Panel (PDP), an inorganic Electroluminescence (EL) panel, or an organic EL panel. The display 14 may also have a touch panel superimposed on it, for example, to detect touch operations by the user's finger. By superimposing a touch panel on the display 14, the smartphone 100 can provide the user with an intuitive operating environment.

[0023] The connection part 15 is an interface for connecting the cuff 16 to the blood pressure monitor body 10. The cuff 16 is connected to the blood pressure monitor body 10 via the connection part 15. The blood pressure monitor body 10 controls the compression of the area to be measured using the cuff 16 via the connection part 15.

[0024] The cuff 16 is a component that is wrapped around the area to be measured. The cuff 16 is made of a flexible material such as cloth. The cuff 16 is equipped with a cuff pressure sensor 17 for detecting cuff pressure. Cuff pressure includes, for example, the pressure exerted by the cuff 16 on the area to be measured and pressure vibrations caused by the pulsation of blood flow in the arteries of the area to be measured.

[0025] <Processing Block of Blood Pressure Monitor Body 10> Figure 2 shows an example of the processing block of the blood pressure monitor body 10 according to the embodiment. The blood pressure monitor body 10 includes an acquisition unit 101, a control unit 102, a generation unit 103, a correction unit 104, a correction model 105, an estimation unit 106, and an output unit 107. The blood pressure monitor body 10 performs processing as each of its parts, such as the acquisition unit 101, control unit 102, generation unit 103, correction unit 104, correction model 105, estimation unit 106, and output unit 107, by having the CPU 11 execute a computer program that has been loaded into the main memory unit 12 in an executable manner.

[0026] The acquisition unit 101 gradually pressurizes the measurement site with the cuff 16 and acquires the time-series change of cuff pressure detected by the cuff pressure sensor 17. The acquisition unit 101 also acquires the time-series change of the pressure pulse wave from the acquired time-series change of cuff pressure. The time-series change of the pressure pulse wave is, for example, an OscilloMetric Waveform (OMW).

[0027] The control unit 102 acquires the envelope for the time-series change of the pressure pulse wave acquired by the acquisition unit 101. When the generated envelope satisfies a predetermined pressure stop condition, the control unit 102 stops applying pressure to the measurement site with the cuff 16 and stops the acquisition of the time-series change of cuff pressure and the time-series change of pressure pulse wave by the acquisition unit 101. The predetermined pressure stop condition is, for example, the timing when the envelope of the pressure pulse wave acquired by the acquisition unit 101 crosses its peak. The envelope of the pressure pulse wave is, for example, a curve formed to connect the peak values ​​of the pulse waveform of each heartbeat in the pressure pulse wave signal. The envelope is, for example, an OscilloMetric Waveform Envelope (OMWE).

[0028] The generation unit 103 generates a mathematical model of the predicted envelope for the pressure pulse wave that would be acquired by the acquisition unit 101 if the control unit 102 had not stopped pressurizing, i.e., if pressurizing by the cuff 16 had continued, based on the pressure pulse wave acquired by the acquisition unit 101 before the control unit 102 stopped pressurizing.

[0029] Figure 3 schematically shows the generation of a mathematical model of the predicted envelope by the generation unit 103 in an embodiment. In Figure 3, it is assumed that the pressurization of the measurement site by the cuff 16 is stopped at time T0. The envelope E1 of the pressure pulse wave P1 up to time T0 is acquired, for example, by the control unit 102. The envelope E1 is formed, for example, by connecting the positive peaks of the positive and negative peaks of the pressure pulse wave P1. Various known techniques can be used to acquire the envelope E1.

[0030] The generation unit 103 fits the envelope E1 with a Gaussian function represented by the following (Equation 1). In (Equation 1), the variable x is the cuff pressure, and A, B, and C are parameters that define the shape of the Gaussian function. The generation unit 103 can obtain a mathematical model of the predicted envelope E2 by fitting the Gaussian function, for example. The predicted envelope E2 is the envelope that is expected to be obtained if the acquisition of the pressure pulse wave continues after time T0. The mathematical model of the predicted envelope E2 is represented by (Equation 1), for example, in which parameters A, B, and C are determined by fitting the Gaussian function.

[0031] Incidentally, the pressure level index α of (Equation 2) can also be used as the predetermined pressure stop condition used by the control unit 102. The pressure level index α is defined as shown in (Equation 2) below, using the parameters B and C of the Gaussian function obtained by fitting. This (Equation 2) defines the relationship (conversion) between the pressure level index α and the cuff pressure x. The pressure level index α is an index that represents the position relative to the envelope.

[0032] Figure 4 shows the relationship between the value of the index α (Equation 2) and the amplitude of the envelope. The index α takes the value of 0 (first value) at the peak position where the amplitude of the envelope is at its maximum, and the absolute value of the index α takes the value of (2log2)1 / 2 (second value) at the position where the amplitude of the envelope is half of the maximum value. Here, log is the natural logarithm.

[0033] Statistically, it is known that the position where the envelope amplitude is half of its maximum value roughly corresponds to DBP and SBP. Therefore, when using the index α, P = -(2log2) １／２The position of approximately points to the position of DBP, and P = (2 log 2) １／２ The position of the index α will roughly indicate the position of the SBP. Based on this knowledge, the pressurization stop threshold Pth may be determined as a predetermined pressurization stop condition such that the index α indicates the position where the peak of the envelope has crossed. That is, when the index α becomes Pth or greater, the control unit 102 may stop pressurizing the area to be measured by the cuff 16 and stop acquiring the time-series changes of the cuff pressure and the time-series changes of the pressure pulse wave by the acquisition unit 101.

[0034] The correction unit 104 corrects the mathematical model of the predicted envelope E2 estimated by the generation unit 103. The correction unit 104 inputs, for example, the feature quantities of the pressure pulse wave P1 into the correction model 105 and obtains corrected values ​​for each parameter A, B, and C. The correction model 105 is a learning model that, when the feature quantities of the pressure pulse wave P1 are input by the correction unit 104, outputs the error of each parameter A, B, and C of the mathematical model of the predicted envelope E2 to each parameter A, B, and C of the mathematical model of the envelope in the case where the acquisition of time-series changes in cuff pressure continues even after a predetermined pressurization stop condition is met. The correction unit 104 applies the corrected values ​​of each parameter A, B, and C obtained from the correction model 105 to each parameter A, B, and C of the mathematical model of the predicted envelope E2 and obtains a corrected mathematical model. For example, if the corrected parameters of parameters B and C are B1 and C1 respectively, then SBP and DBP can be expressed by the following (Equation 3) and (Equation 4). The correction unit 104 may, for example, remove pulse wave information acquired under abnormal conditions, such as during body movement or when irregular pulse waves are generated, or perform interpolation using preceding and succeeding pulse wave information, before extracting the feature quantities. The error output by the correction model 105 is an example of "information related to the parameters of the second mathematical model".

[0035] The features extracted by the correction unit 104 can include, for example, the following information from (1) to (33). The information from (1) to (19) can be said to be information related to the waveform shape of OMW or OMWE. The information from (20) to (25) can be said to be information related to cuff pressure. (1) Minimum value, average value, minimum value. (2) Value at time X% (X is a number from 0 to 100) on the time axis. (3) Time when the value is maximum. (4) Area of ​​the region enclosed by the OMW or OMWE waveform and the time axis. (5) Kurtosis and skewness of the OMW or OMWE waveform. (6) Number of peaks in the OMW or OMWE waveform. (7) Complexity of the OMW or OMWE waveform (for example, sum of squares of the difference between time series data). (8) The number of points in the OMW or OMWE waveform that exceed the average value up to a predetermined point in time (for example, 1 / 3 of the way from the start). (9) The gradient in each of the regions into which the OMW or OMWE waveform is divided (for example, "4"). (10) The value obtained by dividing the time before the peak of the OMW or OMWE waveform by the time after the peak. (11) The rate of change in the height of the OMW or OWME waveform. (12) The area ratio of the waveform to the circumscribing rectangle for each of the regions into which the OMW or OMWE waveform is divided (for example, "4"). (13) The height (amplitude) of the OMW or OWME. (14) Anomalies in the OMW or OWME. (15) The position of the peak of the OMW or OWME. (16) The density of the peaks of the OMW or OWME. (17) Bimodality of OMW and OWME. (18) The value obtained by subtracting the peak time of the lower envelope of OMW from the peak time of the upper envelope of OMW. (19) The area ratio of the area of ​​the upper envelope of OMW to the area of ​​the lower envelope of OMW. (20) Cuff pressure at the OMWE peak. (21) Cuff pressure at the point where the first derivative of OMWE is maximum, and cuff pressure at the point where it is minimum. (22) Cuff pressure at a point in the first half of the OMWE peak that is N times (N is an integer greater than 0 and less than 1) below the peak. (23) Cuff pressure at a point in the second half of the OMWE peak that is N times (N is an integer greater than 0 and less than 1) below the peak. (24) Time from the maximum value of the first derivative of OMWE to the peak. (25) Time from the peak of OMWE to the minimum value of the first derivative. (26) Time length of OMWE.(27) The time to the peak of OMWE divided by the duration of OMWE. (28) The position of the average "-σ" when OMWE is Gaussian fitted. (29) The position of the average "+σ" when OMWE is Gaussian fitted. (30) The value of OMWE at the position of the average "-σ" when OMWE is Gaussian fitted. (31) The value of OMWE at the position of the average "+σ" when OMWE is Gaussian fitted. (32) The maximum gradient of OMWE. (33) The minimum gradient of OMWE.

[0036] We will now further explain some of the features listed in (1) through (33) above. We will now explain the case in which (5) above is adopted as a feature. When (5) above is adopted as a feature, the correction unit 104 may calculate, for example, at least one of the kurtosis and skewness of the pulse wave P1 and input it to the correction model 105 as a feature. The kurtosis varies depending on the size of the peak and the spread of the tail of the pulse wave P1. That is, kurtosis can be said to be a feature related to the size and spread of the peak of the pulse wave P1. In addition, depending on the skewness of the pulse wave P1, the peak position of the pulse wave P1 will shift in the positive direction (to the right when facing the paper) or the negative direction (to the left when facing the paper) in the time axis direction. The pulse wave P1 will have a shape that is distorted to the left or right depending on the skewness. That is, skewness can be said to be a feature related to the position of the peak of the pulse wave P1. The correction unit 104 may use at least one of the kurtosis and skewness of the pressure pulse wave P1 as a feature quantity to be input to the correction model 105.

[0037] The case in which (11) above is adopted as a feature will be explained. Figure 5 is a diagram illustrating the feature of (11) above in the embodiment. In Figure 5, the pressure pulse wave P1 is schematically represented by a set of triangles. The correction unit 104 divides the pressure pulse wave P1 into the first half of the peak and the second half of the peak, centered on the peak of the pressure pulse wave P1. Furthermore, the correction unit 104 divides each of the first half of the peak and the second half of the peak into 1 / 2. The correction unit 104 then calculates the width in the time axis direction (IFX1, IFX2, IFX3, IFX4) and the peak height (IFY1, IFY2, IFY3, IFY4) in each divided region R1, R2, R3, and R4. Here, the peak height is the height difference (relative height) with the peak of the adjacent region, but the peak height may also be the absolute height with pressure 0 as the reference. The correction unit 104 then calculates the gradient of each region using the calculated width and height. For example, the gradient IFG1 of region R1 is calculated as "IFG1 = IFY1 / IFX1". Similarly, the gradient IFG2 of region R2 is calculated as "IFG2 = IFY2 / IFX2", the gradient IFG3 of region R3 is calculated as "IFG3 = IFY3 / IFX3", and the gradient IFG4 of region R4 is calculated as "IFG4 = IFY4 / IFX4". The correction unit 104 may also calculate the ratio of the widths of each calculated region. For example, the correction unit 104 may calculate "IFXR = IFX2 / (IFX2 + IFX3)" as the ratio of the width of region R2 to the sum of the widths of region R2 and region R3. The correction unit 104 may also calculate the ratio of the heights of each calculated region. The correction unit 104 may calculate, for example, "IFYR = IFY2 / (IFY1 + IFY2)" as the ratio of the height of region R2 to the sum of the heights of region R1 and region R2. The correction unit 104 may use the width, gradient, width ratio, and height ratio of each region as feature quantities to be input to the correction model 105.

[0038] The case in which (12) above is adopted as a feature will be explained. Figure 6 is a diagram illustrating the feature in (12) above in the embodiment. The correction unit 104 divides the envelope E1 into a predetermined number of parts (four in the example of Figure 6) in the time axis direction and sets circumscribing rectangles K1, K2, K3, and K4 surrounding the envelope E1. The circumscribing rectangles K1, K2, K3, and K4 are rectangles that are circumscribing to the peak of the pressure pulse wave P1 in each of the divided regions of the envelope E1 and are set in regions where the pressure is 0 or greater. The correction unit 104 then calculates the areas SK1, SK2, SK3, and SK4 of each of the circumscribing rectangles K1, K2, K3, and K4. The correction unit 104 also calculates the areas SC1, SC2, SC3, and SC4 of the regions formed by the envelope E1 and the time axis in each of the divided regions of the envelope E1. Then, the area ratios SK1 / SC1, SK2 / SC2, SK3 / SC3, and SK4 / SC4 in each of the calculated regions are calculated. The correction unit 104 may use the calculated area ratios as feature quantities to be input to the correction model 105.

[0039] The case in which (14) above is adopted as a feature will be explained. An outlier is, for example, a peak caused by the influence of spike noise. Such peaks often appear in the first half of the pulse wave P1. When (14) above is adopted as a feature, the correction unit 104 calculates, for example, the overall average value of the peak values ​​in each heartbeat of the pulse wave P1, and calculates the number of peaks (also called abnormal peaks) that exceed the overall average value in the first half of the pulse wave P1. The correction unit 104 may use the calculated number of abnormal peaks as a feature to be input to the correction model 105.

[0040] The case in which (17) above is adopted as a feature will be explained. When (17) above is adopted as a feature, the correction unit 104 calculates the number of peaks in the envelope E1. The correction unit 104 may use the calculated number of peaks as a feature to be input to the correction model 105.

[0041] A case where the above (19) is adopted as a feature amount will be described. FIG. 7 is a diagram for explaining the feature amount of (19) in the embodiment. When the above (19) is adopted as a feature amount, in addition to the envelope E1 which is an upper envelope, the correction unit 104 acquires an envelope E3 which is a lower envelope formed so as to connect the negative peaks of the pressure pulse wave P1. Then, the correction unit 104 calculates, in a section from time T1 when the pressure pulse wave P1 is acquired to time T0, an area SR1 of a region formed by the envelope E1 and an axis of zero pressure (time axis), and an area SR2 of a region formed by the envelope E3 and the axis of zero pressure (time axis). The correction unit 104 may use the calculated areas SR1 and SR2 as feature quantities to be input to the correction model 105. Alternatively, the correction unit 104 may calculate "SR1 / SR2" which is the area ratio of the area SR1 to the area SR2, and use the calculated area ratio as a feature quantity to be input to the correction model 105.

[0042] Then, the correction unit 104 inputs each parameter of the mathematical model representing the predicted envelope E2 and the feature amounts described above to the correction model 105, and corrects the mathematical model of the predicted envelope E2 generated by the generation unit 103. FIG. 8 is a diagram schematically showing correction of the mathematical model representing the predicted envelope E2 by the correction unit 104 according to the embodiment. Correction of the mathematical model representing the predicted envelope E2 by the correction unit 104 will be described with reference to FIG. 8.

[0043] The generation unit 103 acquires the envelope E1 based on the pressure pulse wave acquired by the acquisition unit 101 before being stopped by the control unit 102 (step M1). The generation unit 103 generates a mathematical model of the predicted envelope E2 based on the envelope E1 (step M2). The correction unit 104 acquires each parameter of the mathematical model of the predicted envelope E2 generated in step M2 (step M3). The correction unit 104 acquires the feature quantities of the pressure pulse wave acquired in step M1 (step M4). The correction unit 104 inputs the parameters acquired in step M3 and the feature quantities acquired in step M4 into the correction model 105 and acquires the error of each parameter of the mathematical model of the predicted envelope E2 generated in step M2 from the correction model 105 (step M5). The correction unit 104 corrects the mathematical model of the predicted envelope E2 generated in step M2 using the error acquired in step M5 (step M6). The estimation unit 106 estimates the blood pressure value using the mathematical model corrected in step M6. The output unit 107 outputs the blood pressure value estimated by the estimation unit 106 to the display 14 (step M7).

[0044] Here, we will briefly explain how the correction model 105 used for correction by the correction unit 104 is constructed. The correction model 105 is constructed by machine learning, for example, using the parameters of the mathematical model of the predicted envelope E2 (at least one of A, B, and C) and at least one of the features described above as explanatory variables, and the error between the mathematical model of the envelope and the parameters of the mathematical model of the predicted envelope E2 when the acquisition of time-series changes in cuff pressure continues even after the predetermined pressurization stop condition is met as the objective variable.

[0045] FIG. 9 is a diagram schematically illustrating the construction of a correction model 105 according to an embodiment. Construction of the correction model 105 will be described with reference to FIG. 9. Although a plurality of subjects are prepared in the present embodiment, the number of subjects may be one. A generation unit 103 fits a Gaussian function to pressure pulse waves acquired even after an envelope for a time-series change in pressure pulse waves satisfies a predetermined pressure stop condition, thereby generating a mathematical model of an envelope (hereinafter referred to as an actually measured envelope E11) for a time-series change in pressure pulse waves acquired even after the predetermined pressure stop condition is satisfied. The generation unit 103 acquires parameters of the mathematical model of the actually measured envelope E11 generated in step T1. The generation unit 103 acquires an envelope E1 based on pressure pulse waves acquired by an acquisition unit 101 before the acquisition is stopped by a control unit 102 (step T3). The generation unit 103 generates a mathematical model of a predicted envelope E2 based on the envelope E1 (step T4). The generation unit 103 acquires parameters of the mathematical model of the predicted envelope E2 generated in step T4 (step T5). The generation unit 103 acquires an error of the parameters acquired in step T2 with respect to the parameters acquired in step T5 (step T6). A correction unit 104 calculates a feature quantity of the pressure pulse wave acquired in step T3 (step T7). Machine learning is performed using the feature quantity extracted in step T7 as an explanatory variable and the error calculated in step T6 as an objective variable (step T8). The correction model 105 is constructed through the machine learning in step T8.

[0046] The processing from step T1 to step T8 is executed for each of subjects H1 to HN. That is, the correction model 105 is not a learning model that improves blood pressure measurement accuracy for a single subject, but can be said to be a model that improves blood pressure value measurement accuracy by using measurement results from a plurality of subjects. Note that the correction model 105 is not limited to one constructed by machine learning, and a statistical model such as a multiple regression model may be used, for example.

[0047] Returning to FIG. 2, an estimation unit 106 estimates a blood pressure value using the mathematical model corrected by the correction unit 104.

[0048] The output unit 107 outputs the blood pressure value estimated by the estimation unit 106 using the parameter-corrected mathematical model to, for example, the display 14.

[0049] <Processing Flow> Figure 10 shows an example of the processing flow for blood pressure measurement by the blood pressure monitor 1 according to this embodiment. Hereinafter, an example of the processing flow for blood pressure measurement by the blood pressure monitor 1 will be described with reference to Figure 10.

[0050] In step S1, the acquisition unit 101 acquires the pressure pulse wave while gradually applying pressure to the measurement site with the cuff 16.

[0051] In step S2, the control unit 102 acquires the envelope for the time-series change of the pressure pulse wave acquired in step S2. If the acquired envelope satisfies the predetermined pressure stop condition (satisfied in step S3), the process proceeds to step S4. If the acquired envelope does not satisfy the predetermined pressure stop condition (not satisfied in step S3), the process proceeds to step S1.

[0052] In step S4, the control unit 102 stops pressurizing the cuff 16 and stops acquiring the pressure pulse wave by the acquisition unit 101. Once pressurizing is stopped in step S4, the process proceeds to steps S5 and S6.

[0053] In step S5, the generation unit 103 generates a predicted envelope E2 based on the envelope E1 for the pressure pulse wave acquired up to step S4.

[0054] In step S6, the correction unit 104 acquires the characteristic values ​​of the pressure pulse wave acquired up to the time the pressurization was stopped in step S4.

[0055] In step S7, the correction unit 104 inputs the feature quantities acquired in step S6 into the correction model 105 to obtain correction values ​​for the parameters of the mathematical model of the envelope E1. The correction unit 104 uses the obtained correction values ​​to correct the mathematical model of the envelope E1.

[0056] In step S8, the estimation unit 106 estimates the blood pressure value using the mathematical model corrected in step S7.

[0057] In step S9, the output unit 107 outputs the blood pressure value estimated in step S8 to, for example, the display 14.

[0058] According to this embodiment, the mathematical model of the envelope E1 generated by the generation unit 103 is corrected by the correction unit 104, thereby increasing the accuracy of blood pressure value estimation as much as possible. In other words, in this embodiment, errors that may be included in the mathematical model of the predicted envelope E2 can be corrected, thereby increasing the accuracy of blood pressure value estimation using the mathematical model.

[0059] In this embodiment, the predetermined pressure stop condition is, for example, the timing when the envelope E1 exceeds its highest value. When the envelope E1 exceeds its highest value, it is considered that the error between the envelope E1 and its mathematical model becomes relatively small. Therefore, according to this embodiment, both SBP and DBP can be suitably estimated.

[0060] Furthermore, in this embodiment, various features can be used in the correction model 105, as described above as (1) to (33). By appropriately selecting the features to be used in the correction model 105 based on the accuracy of blood pressure estimation, the computational load on the blood pressure monitor body 10, etc., a blood pressure monitor 1 can be provided that suits the user's intended use.

[0061] <Modification> In the embodiment described above, the correction model 105 was constructed to output the errors of each parameter A, B, and C of the mathematical model of the predicted envelope E2 for each parameter A, B, and C of the mathematical model of the measured envelope E11. However, the correction model 105 may be constructed to output the corrected parameters of each parameter A, B, and C of the predicted envelope E2. In such a case, the correction unit 104 can construct the correction model 105 by performing machine learning with the features extracted in step T7 of Figure 9 as explanatory variables and each parameter of the mathematical model of the measured envelope E11 obtained in step T2 of Figure 9 as the target variable. In this case, steps T5 and T6 of Figure 9 may be omitted. The correction unit 104 can then correct the mathematical model of the predicted envelope E2 by replacing the parameters of the mathematical model of the predicted envelope E2 generated in step T4 of Figure 9 with the parameters obtained from the correction model 105.

[0062] In the embodiment described above, the correction model 105 is stored in the auxiliary storage unit 13 within the blood pressure monitor body 10. However, the correction model 105 may be stored in a location other than the auxiliary storage unit 13. For example, the correction model 105 may be stored in a server or cloud system connected to the blood pressure monitor body 10 by a computer network.

[0063] The embodiments and variations disclosed above can be combined in any way.

[0064] <Computer-readable recording medium> An information processing program that enables a computer or other machine or device (hereinafter referred to as "computer, etc.") to perform any of the above functions can be recorded on a computer-readable recording medium. By having the computer, etc. read and execute the program on this recording medium, it can provide that function.

[0065] Here, a recording medium that can be read by a computer refers to a recording medium that stores information such as data and programs through electrical, magnetic, optical, mechanical, or chemical means and can be read by a computer. Examples of such recording media that can be removed from a computer include flexible disks, magneto-optical disks, Compact Disc Read Only Memory (CD-ROM), Compact Disc-Recordable (CD-R), Compact Disc-ReWritable (CD-RW), Digital Versatile Disc (DVD), Blu-ray Disc (BD), Digital Audio Tape (DAT), 8mm tape, flash memory, external hard disk drives, and Solid State Drives (SSDs). Furthermore, there are internal hard disk drives, SSDs, and ROMs as recording media fixed to computers and other devices.

[0066] <Note 1> A sensor (17) for detecting the pressure of a cuff (16) for compressing the part of the person being measured; a pulse wave acquisition unit (101) for acquiring a pulse wave from the output signal of the sensor (17); a control unit (102) for stopping the pressurization when a first pulse wave (P1) acquired while gradually pressurizing the part of the person being measured using the cuff (16) satisfies a predetermined pressurization stop condition; and a generation unit (103) for generating a first mathematical model (E2, equation (1)) of the predicted envelope of a second pulse wave that is expected to be obtained if the pressurization is continued, based on the first pulse wave (P1) acquired up to the point where the predetermined pressurization stop condition is met. A blood pressure monitor (1) comprises: a correction model (105) constructed by machine learning, in which the characteristic quantities related to the first pulse wave (P1) acquired up to the predetermined pressurization stop condition is met for each subject (H1, H2, ..., HN) are used as explanatory variables, and information related to the parameters of a second mathematical model (E11, (Equation 1)) of the measured envelope of the third pulse wave (P2) that is continuously acquired even after the predetermined pressurization stop condition is met is used as the objective variable; a correction unit (104) that acquires the information output by the correction model (105) by inputting the characteristic quantities into the correction model (105), and corrects the first mathematical model (E2, Equation (1)) using the acquired information; an estimation unit (106) that estimates the blood pressure value of the subject using the first mathematical model (E2, Equation (1)) corrected by the correction unit (104); and an output unit (107) that outputs the blood pressure value estimated by the estimation unit (106). <Note 2> The blood pressure monitor according to Note 1, wherein the information includes the error between the parameters (A, B, C of Equation (1)) of the first mathematical model (E2, Equation (1)) and the parameters (A, B, C of Equation (1)) of the second mathematical model (E11, (Equation 1)). <Note 3> The blood pressure monitor according to Note 1, wherein the information includes the parameters (A, B, C of Equation (1)) of the second mathematical model (E11, (Equation 1)). <Note 4> The blood pressure monitor according to any one of Notes 1 to 3, wherein the feature quantity relating to the first pulse wave (P1) includes the amplitude of the first pulse wave (P1).<Note 5> The blood pressure monitor according to any one of Notes 1 to 4, wherein the characteristic quantity relating to the first pulse wave (P1) includes the ratio of the area of ​​the positive region to the area of ​​the negative region of the first pulse wave (P1). <Note 6> The computer performs the following steps: a pulse wave acquisition step of acquiring a pulse wave from the output signal of a sensor that detects the pressure of a cuff used to compress the part of the person being measured; a pressurization step of stopping pressurization when the first pulse wave acquired while gradually pressurizing the part of the person being measured using the cuff satisfies a predetermined pressurization stop condition; a generation step of generating a first mathematical model of the predicted envelope of a second pulse wave that is expected to be obtained if pressurization is continued, based on the first pulse wave acquired up to the point where the predetermined pressurization stop condition is met; a correction step of acquiring the information output by the correction model by inputting the feature quantities into a correction model in which the feature quantities related to the first pulse wave acquired up to the point where the predetermined pressurization stop condition is met are used as explanatory variables and information related to the parameters of the second mathematical model of the measured envelope of a third pulse wave that has been continuously acquired even after the predetermined pressurization stop condition is met is used as the objective variable, and correcting the first mathematical model using the acquired information; an estimation step of estimating the blood pressure value of the person being measured using the first mathematical model corrected in the correction step; and an output step of outputting the blood pressure value estimated in the estimation step. Blood pressure measurement methods.<Note 7> The following steps are performed on a computer: a pulse wave acquisition step of acquiring a pulse wave from the output signal of a sensor that detects the pressure of a cuff used to compress the part of the person being measured; a pressurization step of stopping pressurization when the first pulse wave acquired while gradually pressurizing the part of the person being measured using the cuff satisfies a predetermined pressurization stop condition; a generation step of generating a first mathematical model of the predicted envelope of a second pulse wave that is expected to be obtained if pressurization is continued, based on the first pulse wave acquired up to the point where the predetermined pressurization stop condition is met; a correction step of acquiring the information output by the correction model by inputting the feature quantities into a correction model in which the feature quantities related to the first pulse wave acquired up to the point where the predetermined pressurization stop condition is met are used as explanatory variables and information related to the parameters of the second mathematical model of the measured envelope of a third pulse wave that has been continuously acquired even after the predetermined pressurization stop condition is met is used as the objective variable, and correcting the first mathematical model using the acquired information; an estimation step of estimating the blood pressure value of the person being measured using the first mathematical model corrected in the correction step; and an output step of outputting the blood pressure value estimated in the estimation step. Blood pressure measurement program. <Note 8> A method for constructing a learning model, wherein a computer executes the following steps: a first acquisition step of acquiring feature quantities related to a first pulse wave acquired while gradually pressurizing the part to be measured using a cuff until a predetermined pressurization stop condition is met; a second acquisition step of acquiring information related to the parameters of a second mathematical model of the measured envelope of a third pulse wave, which is continuously acquired even after the predetermined pressurization stop condition is met while gradually pressurizing the part to be measured using the cuff; and a construction step of constructing a learning model by machine learning with the feature quantities as explanatory variables and the information as the target variable.<Note 9> A program for constructing a learning model, which causes a computer to perform the following steps: a first acquisition step of acquiring feature quantities related to a first pulse wave acquired while gradually pressurizing the part to be measured using a cuff until a predetermined pressurization stop condition is met; a second acquisition step of acquiring information related to the parameters of a second mathematical model of the measured envelope of a third pulse wave, which is continuously acquired even after the predetermined pressurization stop condition is met while gradually pressurizing the part to be measured using the cuff; and a construction step of constructing a learning model by machine learning with the feature quantities as explanatory variables and the information as the target variable.

[0067] 1. Blood pressure monitor 10. Blood pressure monitor body 11. CPU 12. Main memory unit 13. Auxiliary memory unit 14. Display 15. Connection unit 16. Cuff 17. Cuff pressure sensor 101. Acquisition unit 102. Control unit 103. Generation unit 104. Correction unit 105. Correction model 106. Estimation unit 107. Output unit B1. Connection bus E1. Envelope E2. Predicted envelope E3. Envelope E11. Measured envelope P1. Pressure pulse wave P2. Pressure pulse wave

Claims

1. The system comprises: a sensor that detects the pressure of a cuff used to compress the part of a person to be measured; a pulse wave acquisition unit that acquires a pulse wave from the output signal of the sensor; a control unit that stops pressurizing when a first pulse wave acquired while gradually pressurizing the part of the person to be measured using the cuff satisfies a predetermined pressurization stop condition; a generation unit that generates a first mathematical model of the predicted envelope of a second pulse wave that is expected to be obtained if pressurization is continued, based on the first pulse wave acquired up to the time the predetermined pressurization stop condition is met; a correction model for a subject that uses feature quantities related to the first pulse wave acquired up to the time the predetermined pressurization stop condition is met as explanatory variables, and information related to the parameters of the second mathematical model of the measured envelope of a third pulse wave, which is continuously acquired even after the predetermined pressurization stop condition is met, as the objective variable; a correction unit that acquires the information output by the correction model by inputting the feature quantities into the correction model, and corrects the first mathematical model using the acquired information; an estimation unit that estimates the blood pressure value of the person to be measured using the first mathematical model corrected by the correction unit; and an output unit that outputs the blood pressure value estimated by the estimation unit. Blood pressure monitor.

2. The blood pressure monitor according to claim 1, wherein the information includes the error of the parameters of the first mathematical model with respect to the parameters of the second mathematical model.

3. The blood pressure monitor according to claim 1, wherein the information includes the parameters of the second mathematical model.

4. The blood pressure monitor according to claim 1, wherein the characteristic quantity relating to the first pulse wave includes the amplitude of the first pulse wave.

5. The blood pressure monitor according to claim 1, wherein the characteristic quantity relating to the first pulse wave includes the ratio of the area of ​​the positive region to the area of ​​the negative region of the first pulse wave.

6. A blood pressure measurement method comprising: a pulse wave acquisition step of acquiring a pulse wave from the output signal of a sensor that detects the pressure of a cuff used to compress the part of a person to be measured; a pressurization step of gradually pressurizing the part of a person to be measured using the cuff and stopping the pressurization when the acquired first pulse wave satisfies a predetermined pressurization stop condition; a generation step of generating a first mathematical model of the predicted envelope of a second pulse wave that is expected to be obtained if the pressurization is continued, based on the first pulse wave acquired up to the point where the predetermined pressurization stop condition is met; a correction step of acquiring the information output by the correction model by inputting the feature quantities into a correction model in which the feature quantities related to the first pulse wave acquired up to the point where the predetermined pressurization stop condition is met are used as explanatory variables and information related to the parameters of the second mathematical model of the measured envelope of a third pulse wave, which has been continuously acquired even after the predetermined pressurization stop condition is met, and correcting the first mathematical model using the acquired information; an estimation step of estimating the blood pressure value of the person to be measured using the first mathematical model corrected in the correction step; and an output step of outputting the blood pressure value estimated in the estimation step.

7. A blood pressure measurement program that causes a computer to execute the following steps: a pulse wave acquisition step of acquiring a pulse wave from the output signal of a sensor that detects the pressure of a cuff used to compress the part of a person to be measured; a pressurization step of stopping pressurization when a first pulse wave acquired while gradually pressurizing the part of the person to be measured using the cuff satisfies a predetermined pressurization stop condition; a generation step of generating a first mathematical model of the predicted envelope of a second pulse wave that is expected to be obtained if pressurization is continued, based on the first pulse wave acquired up to the time the predetermined pressurization stop condition is met; a correction step of acquiring the information output by the correction model by inputting the feature quantities into a correction model in which the feature quantities related to the first pulse wave acquired up to the time the predetermined pressurization stop condition is met for a subject are used as explanatory variables and information related to the parameters of the second mathematical model of the measured envelope of a third pulse wave that has been continuously acquired even after the predetermined pressurization stop condition is met is used as the objective variable, and correcting the first mathematical model using the acquired information; an estimation step of estimating the blood pressure value of the person to be measured using the first mathematical model corrected in the correction step; and an output step of outputting the blood pressure value estimated in the estimation step.

8. A method for constructing a learning model, wherein a computer performs the following steps: a first acquisition step of acquiring feature quantities related to a first pulse wave acquired while gradually pressurizing the part to be measured using a cuff until a predetermined pressurization stop condition is met; a second acquisition step of acquiring information related to the parameters of a second mathematical model of the measured envelope of a third pulse wave, which is continuously acquired even after the predetermined pressurization stop condition is met while gradually pressurizing the part to be measured using the cuff; and a construction step of constructing a learning model by machine learning with the feature quantities as explanatory variables and the information as the target variable.

9. A program for constructing a learning model, which causes a computer to perform the following steps: a first acquisition step of acquiring feature quantities related to a first pulse wave acquired while gradually pressurizing the part to be measured using a cuff until a predetermined pressurization stop condition is met; a second acquisition step of acquiring information related to the parameters of a second mathematical model of the measured envelope of a third pulse wave, which is continuously acquired even after the predetermined pressurization stop condition is met while gradually pressurizing the part to be measured using the cuff; and a construction step of constructing a learning model by machine learning with the feature quantities as explanatory variables and the information as the target variable.

Images